Introduction

Obesity is a major global public health issue that increases the risk of type 2 diabetes, fatty liver disease, hypertension, cardiovascular disease, and multiple cancers1. Alongside urbanization, changes in food supply and eating habits, and a reduction in physical activity, the obesity rate more than tripled in low socio-economic populations such as the Middle East, Oceania, and China2. Currently, more than one-third of the global population is overweight or obese. At least 60% of obese patients live in developing countries, and the prevalence of hypertension and obesity-related cardiac metabolic disorders is rapidly increasing3. To date, there has been insufficient focus on the impact of obesity with hypertension incidence in rural China.

Hypertension is a known variable risk factor for cardiovascular disease, with obesity an independent risk factor for hypertension4. In addition, people with obesity are 3.5 times more likely to suffer from hypertension, and 60% of hypertension cases are attributed to an increase in fat storage. Previous data showed the prevalence of hypertension in obese individuals with a body mass index (BMI) > 30 kg/m2 was 42.5%, significantly higher than 15.3% in non-obese individuals5. The conventional anthropometric indicators that define obesity include BMI, waist circumference (WC)6, waist-height ratio (WHtR), and waist-hip ratio (WHR)7. Unconventional new indicators proposed in recent years include a body shape index (ABSI)8, and body roundness index (BRI)9.

Epidemiological studies evaluated the relationship between conventional anthropometric obesity indices and blood pressure (BP). However, the results are inconsistent and the optimal index unconclusive. Zhang et al. of a recent cross-sectional study concluded that WC and BMI were related to hypertension, and an additive interaction between WC and BMI existed10. A prospective cohort study returned comparable findings11. However, some authors question the predictive value of BMI. It cannot distinguish between fat and muscle mass nor reflect the individual fat distribution12. Some studies suggested WC, WHtR, and WHR were alternative indicators of obesity. However, other studies did not concur with WHR as a predictor of hypertension13. In terms of unconventional obesity indicators14, BRI was found to be able to determine whether cardiovascular disease (CVD) existed, while ABSI cannot. However, the ability of BRI to recognize CVD was not better than BMI and WC. Other studies found ABSI could predict premature death by expressing the additional risk from high WC, using a simple formula8. At present, the predictive value of unconventional obesity-related indicators for hypertension is unelucidated.

Preliminary studies inferred that inconsistency in the research of optimal conventional obesity indicators may be related to nationality, eating habits, climate differences, and baseline BP. Moreover, the relationship between obesity and hypertension may also be affected by sex and age15, yet previous studies were mostly cross-sectional and not cross-grouping by age and sex. The relationship in different sex and age groups in the Chinese population between obesity indicators and the incidence of hypertension is not entirely clear, and the relationship between unconventional obesity indicators (BRI and ABSI) and hypertension has not yet been explored. Compared to in urban areas, rural areas in China have a high incidence rate of hypertension and even cardiovascular diseases. In 2018, the incidence of hypertension in urban areas was 25.7%, while in rural areas it was 29.4%16. Moreover, the relative incidence rate in rural areas was higher, but the control rate was lower. Previous studies had found that the incidence rate of hypertension in rural areas in Northeast China was as high as 50%, but the control rate was only 5.5%17. Secondly, medical care in rural areas was relatively backward18, and the burden of hypertension was relatively greater. Therefore, it is necessary to pay more attention to the prevention of hypertension before its onset19. In addition, the obesity incidence rate in China is currently increasing, and it is expected to increase to 70.5% by 2030. Although the urban–rural gap is narrowing, it is possible to achieve in the future a reversal of this trend of incidence in rural areas surpassing that in urban areas20. Therefore, to address this urgent issue that China needs to resolve, this study aimed to explore whether unconventional and conventional obesity indicators can predict new-onset hypertension in a prospective cohort in rural areas of Northeast China. We also aimed to compare the predictive value of various obesity-related anthropometric indices for new-onset hypertension prediction among different sex and age groups.

Materials and methods

Study design and data collection

The Northeast China Rural Cardiovascular Health Study (NCRCHS) was conducted from January 2012 to August 2013. The study design of this community-based prospective cohort study was previously described21. First, three counties (Dawa, Zhangwu, and Liaoyang) were selected from the eastern, southern, and northern regions of Liaoning Province. In the second stage, one town was randomly selected from each county (three towns total). In the third stage, 8–10 rural villages were randomly selected from each town (26 in total). Participants who were pregnant, had malignancies, or had mental disorders were excluded. In total, 11,956 permanent residents aged ≥ 35 years in each village were invited to participate. The response rate was 89.5%. Overall, 10,700 participants agreed and qualified to participate in our follow-up study, and baseline information on each participant was collected.

In this study, trained researchers from China Medical University interviewed each participant using a standardized questionnaire, face-to-face interview, physical examination, and blood biochemical test. Clinical, demographic, and lifestyle information of all participants was collected. Exclusions included 97 people due to incomplete baseline BP data, 5548 people with baseline hypertension, 48 deaths in the median follow-up of 2.59 years, and 884 people without follow-up BP data information. Finally, we collected data from 4123 participants, including 2377 women, to assess the correlation between BMI, WC, WHR, WHtR, BRI, ABSI, and new-onset hypertension (Fig. 1).

Figure 1
figure 1

Flow chart of subject inclusion and exclusion.

Full size image

Exposure and outcome ascertainment

Anthropometric indicators were measured according to standard procedures. We used a standard rangefinder to measure the barefoot height of participants. The subjects wore light indoor clothes when using the electronic scale to measure weight. The WC was measured with an inelastic anthropometric belt at about belly level at the midpoint between the lowest edge of the ribs and the level of the anterior superior iliac crest. We measured twice at the end of normal exhalation and recorded the average value. The hip circumference was measured at the maximum protrusion of the gluteal muscle. Readings were recorded to the nearest 0.1 cm and 0.1 kg. BMI was calculated by dividing an individual's weight in kilograms by the square of their height in meters. Considering the small stature of subjects in rural China, according to the criteria of the China Obesity Working Group (divided into < 18.5, 18.5–23.9, 24.0–27.9, and ≥ 28 kg/m2), a BMI > 28 kg/m2 was defined as obesity22. WHR was calculated by dividing WC by hip circumference, and WHtR was calculated by dividing WC by height. According to World Health Organization standards, the central obesity of men was WC ≥ 90 cm or WHtR ≥ 0.5 or WHR ≥ 0.9; that of women was WC ≥ 80 cm or WHtR ≥ 0.5, or WHR ≥ 0.857. ABSI and BRI were calculated as follows8,9:

$${\text{ABSI }} = {\text{ WC}}\div(BMI^{2/3}\times Height^{1/2})$$
$${\text{BRI }} = {364}.{2} - {365}.{5 } \times \sqrt {1 - (WC^{2} \div (\pi \times {\text{Height}})^{2} )} .$$

Specific consulting rooms were used at the recruitment and follow-up sites. During BP measurement, researchers from the project team used HEM-907, a standard electronic blood pressure monitor (Omron, Japan), to evaluate the right arm BP after participants had rested in a sitting position for 10 min. Each measurement interval was at least 30 s. The average of the three closest readings was taken as the BP value. All participants were instructed not to exercise, smoke, or drink irritant-containing beverages (e.g., tea, coffee, alcohol) before BP measurement. Participants also underwent detailed cardiovascular examinations in 2015 which included face-to-face visits and on-site blood pressure measurement review. Participants were also required to report on blood pressure measured at home or village clinics and on related diagnostic and treatment processes. The standard parameters for defining hypertension were a systolic blood pressure ≥ 140 mmHg and/or a diastolic blood pressure ≥ 90 mmHg, or the use of anti-hypertension drugs23.

Covariates

Information on demographic characteristics (e.g., age, sex, marital status, education level, income level) and behavioral measures (smoking, exercise, drinking, physical activity intensity) was obtained through standardized questionnaires, and the measures were carried out in strict accordance with the national standard for basic public health services (2011). All participants were required to collect pre-elbow venous blood on an empty stomach (more than 12 h) in the morning. After it stood at room temperature for half an hour, the blood sample was centrifuged at 3000 rpm for 5 min. The supernatant was collected and stored at − 20 °C and transferred to the laboratory within 4 h for immediate testing. The Abbott Diagnostics C800i automatic analyzer (Abbott Laboratories, Abbott Park, IL, USA) was used for routine index determination. Blood lipid abnormalities were classified according to the NCEPATPIII standards and liver function cutoff values according to previous literature standards24,25: high low-density lipoprotein cholesterol (LDL-C) was defined as ≥ 4.14 mmol/L; low high-density lipoprotein cholesterol (HDL-C) was defined as < 1.04 mmol/L; high total cholesterol (TC) was defined as ≥ 6.22 mmol/L; and high triglyceride (TG) was defined as ≥ 2.27 mmol/L. Increased aspartate transaminase (AST) was defined as > 35 U/L; high alanine transaminase (ALT) was defined as > 40 U/L; high uric acid (UA) for males was ≥ 420 μmol/L and for females was ≥ 360 μmol/L26; and high white blood cell (WBC) count was defined based on the laboratory reference cutoff as greater than 10 × 109.

Statistical analysis

The Shapiro–Wilk and Kolmogorov–Smirnov tests were used to test the normality of the dataset. All continuous variables are described as mean ± standard deviation (SD) after the normality test, while categorical variables are described as percentages. The Student’s t-test and Chi-square test were used to estimate the differences between continuous variables and categorical variables in different groups. This study estimated the following obesity-related variables: BMI, WC, WHR, WHtR, BRI, and ABSI. Hazard ratios (HRs) and 95% confidence intervals (CIs) were used to estimate the risk of new-onset hypertension for each obesity index (all binary variables) with the help of the Cox regression model. For each index, the area under the receiver operating characteristic (ROC) curve (AUC) was calculated, and the clinical decision curve of each index was drawn. The clinical usefulness of the index was evaluated by decision curve analysis (DCA), which can quantify the net benefits at different threshold probabilities. The least absolute shrinkage and selection operator (LASSO) strategy (variables with a Wald Chi-square test result of P < 0.2 in the univariate Cox proportional hazards regression model were used as candidate factors), an efficient statistical method for coping with high-dimensional data27, was employed to screen the most useful predictive risk factors in Model 2 and the hypertension model. The optimal filtering of variables in the LASSO regression model was determined using tenfold cross-validation and the minimum criteria or one standard error of the minimum criteria (1SE criteria). Finally, Harrell’s C-index and the AUC of the prediction model combined with other risk factors were calculated in different sex and age groups (60 years was the boundary), and the maximum Youden index (sensitivity + specificity − 1) was used to explore the best prediction index of new-onset hypertension in different subgroups. Meanwhile, the area under the ROC curve (AUC) between 0.5 and 0.6 suggests poor accuracy of the diagnostic test. AUC between 0.6 and 0.7 suggests sufficient accuracy, between 0.7 and 0.8 good accuracy, and between 0.8 and 0.9 very good accuracy, whereas AUC higher than 0.9 suggests excellent accuracy of the diagnostic test28. The method used for missing data in other non-main research variables was multiple imputation. All tests were two-sided, and P-values < 0.05 were considered statistically significant. SPSS version 23.0 (IBM Corp., Armonk, NY, USA) and R software (version 3.6.3; R Foundation, Vienna, Austria) were used for the analyses.

Patient and public involvement statement

The patients were not actively involved during the design and conduct of this study. The patients and the general public will be informed of the study results through peer-reviewed journals.

Patient consent for publication

Consent obtained directly from patient(s).

Ethical approval

This study was conducted according to the Helsinki Declaration of the World Medical Association and was approved by the Ethics Committee of China Medical University (Shenyang, China; AF-SDP-7-1, 0-01). All participants provided written informed consent.

Results

Baseline characteristics and Cox regression analysis

The baseline demographic characteristics and the results of the univariate Cox regression analysis of the 4123 eligible participants are shown in Table 1. The average age was 50.11 ± 9.40 years. Naturally, the cumulative incidence rate of new-onset hypertension increased with age (the rates for age < 40 and 40–60 were 12.39% and 19.33%, respectively). The highest incidence of new-onset hypertension was 28.92% (188/650) in people > 60 years old, and the risk of hypertension increased 1.84 times (95% CI 1.42–2.37) and 3.69 times (95% CI 2.78–4.91) when reaching 40 and 60 years old, respectively.

Table 1 Baseline characteristics and univariate cox regression analysis.
Full size table

There was a significant difference in the baseline BP between the two groups with or without new-onset hypertension (P < 0.001). Male participants accounted for 42.34%, and men were more prone to developing high blood pressure than women (24.23% vs. 16.62%). Participants with lower education levels (junior high school and below) accounted for the majority (89.69%) of those with new-onset hypertension. There were differences in the education level and marital status between the two groups. Factors such as drinking status, smoking status, baseline BP, FBG, diabetes, snoring, family history of hypertension or stroke, low-frequency intake of bean products, hyperuricemia, hypercholesterolemia, and estimated glomerular filtration rate levels were significantly associated with the risk of new-onset hypertension, with statistical differences between the two groups.

Obesity-related anthropometric characteristics

BRI, ABSI, BMI, WHR, WC and WHtR were statistically different between participants with and without new-onset hypertension in the whole population (data not shown). The participants were divided into four subgroups according to the two variables of age (60 years old) and sex. Table 2 shows the characteristics of obesity-related anthropometric indicators of participants according to sex and age. There was no significant difference between the obesity indexes of the two groups (P > 0.05) among women aged > 60 when the indices were analyzed as continuous variables. Among older men, the difference was not significant for BMI and ABSI between the group who progressed to hypertension and the non-hypertension group (P > 005). Individuals with hypertension had higher BMI, BRI, WHtR, ABSI, and WHR than those with no high blood pressure, regardless of sex.

Table 2 Characteristics of obesity-related anthropometric indicators of participants according to sex and age.
Full size table

Clinical usefulness of anthropometric indices related to obesity

The ROC curve (Fig. 2) shows the preliminary exploration of the predicted values of obesity-related anthropometric indices for new-onset hypertension. The curve suggests that WHR and WC have a better predictive value, with AUCs of 0.605 and 0.598, while the AUCs of ABSI and BRI were 0.574 and 0.585, respectively, which were not better than the conventional WHR. Moreover, the predictive ability of BMI was the poorest, with an AUC of 0.566. Figure 3 shows the DCA curves of new-onset hypertension within 2.5 years, suggesting that both conventional and unconventional obesity indices can improve the identification of new-onset hypertension. In terms of hypertension occurrence in the general population within the prediction threshold range of 15–25%, the prediction of BP intervention management according to both conventional and unconventional obesity indices was better than the “treat-all-patients” or “no-treatment” approaches. Overall, the standardized net benefit of the WHR was the greatest. Considering that the incidence of hypertension in the cohort over the 2.5 years was 19.84%, we suggest a threshold probability of 20% for the participants. The DCA curve also shows that WHR and WC may have the best potential predictive benefits.

Figure 2
figure 2

Area under the receiver operating characteristic (ROC) curve (AUC) of each obesity index in new-onset hypertension. WHR waist hip ratio, WC waist circumference, BRI body roundness index, WHtR waist height ratio, ABSI a body shape index, BMI body mass index.

Full size image
Figure 3
figure 3

Decision curve analysis for the obesity indices. The x- and y-axes measure the threshold probability and standardized net benefit, respectively. The horizontal black line represents the hypothesis that none of the general population will receive an intervention. The grey solid line represents the hypothesis that interventions are offered to all participants. The other 6 lines represent the net benefits of offering interventions according to the threshold probabilities derived from the 6 obesity indices. WHtR waist height ratio, WC waist circumference, BMI body mass index, WHR waist hip ratio, BRI body roundness index, ABSI a body shape index.

Full size image

Subgroup analysis

The relationships between obesity-related indices and the risk of new-onset hypertension are shown in Table 3 (subgroup analysis based on sex and age). In the Cox regression analysis, Model 1 was a crude model without any adjustment, while Model 2 was adjusted for risk factors, including SBP, DBP, pulse, diabetes, family history of hypertension, family history of stroke, intake frequency of the beans, physical activity intensity, hyperuricemia, hypertriglyceridemia, and snoring. Among the men ≤ 60 years, WHtR, WHR, ABSI, and BRI were associated with an increased risk of hypertension (even after adjustment for other risk factors in Model 2). WHR was associated with the highest risk of hypertension with an HR of 2.28 (95% CI 1.73–3.01), while WC in men > 60 years was associated with the highest risk of hypertension, and the HR was 3.24 (95% CI 1.48–7.08). Regardless of age, WC, WHtR, WHR, ABSI, and BRI were positively correlated with the risk of hypertension in women. The HRs of WHR were the highest in all age groups (HR 2.38 in the ≤ 60 years age group; HR: 2.51 in the > 60 years age group). However, BMI did not predict the risk of new-onset hypertension regardless of sex and age in Model 2.

Table 3 Subgroup analysis of the risk of new-onset hypertension.
Full size table

Comparison of predicted values

Table 4 shows the comparison of predicted values of obesity indicators based on different sex and age groups. We compared the AUCs and C-indices of the hypertension prediction models constructed by obesity-related indices with age, SBP, DBP, pulse, diabetes, family history of hypertension, family history of stroke, intake frequency of the beans, physical activity intensity, hyperuricemia, hypertriglyceridemia, and snoring. To avoid over-filling the constructed prediction models and possible collinearity between variables, the screening of risk factors included the variables screened by LASSO analysis and conventional independent predictors of hypertension. Among the obesity-related anthropometric indices of women in the two age groups, WHR was accompanied by the largest AUCs (0.793 in the ≤ 60 years age group and 0.716 in the > 60 years age group) and C-indices (0.776 and 0.696, respectively). In addition, WC (AUC = 0.759, C-index = 0.746) and WHR (AUC = 0.788, C-index = 0.721) were the best predictors of new-onset hypertension in men ≤ 60 and > 60 years, respectively.

Table 4 Comparison of the predicted value of obesity indicators based on different sex and age groups.
Full size table

Discussion

In this study, we supplemented the predictive values of the unconventional obesity-related indices ABSI and BRI when forecasting hypertension. At the same time, we compared the predictive values of obesity-related anthropometric indices among Chinese adults divided into four groups according to sex and age for predicting the incidence of hypertension, as well as exploring the best predictive indicators. By using the survey data of a prospective cohort of general people in rural areas of northeast China, we found strong evidence of a positive correlation between obesity-related anthropometric indicators and the risk of hypertension. During the median follow-up of 2.59 years, 818 (19.8%) new cases of hypertension occurred. Unconventional obesity indices BRI and ABSI had predictive value for new-onset hypertension, but they were not better than conventional indices, and BMI had limited predictive value. The best predictor of hypertension in women from the two age groups was WHR, while WC and WHR were the best predictors of new-onset hypertension in men ≤ 60 and > 60 years, respectively. Our results highlight the obesity index is most closely related to new-onset hypertension in the short term in different sex and age groups. This will benefit primary health care regarding the control of and monitoring of the incidence of hypertension.

We also found that compared with other obesity indices, WHR and WC have a greater impact on the risk of hypertension in the general population. ABSI and BRI also have some predictive values as unconventional obesity anthropometric indicators. These findings are consistent with previous ultrasound studies, which found that visceral fat increases the risk of hypertension29. Excessive visceral fat distribution is often accompanied by changes in inflammation levels, hormones, and endothelial cells (including decreased adipocytokines, increased free fatty acids, uric acid, leptin, and vascular endothelial growth factor)30. Obesity can raise BP, and is reported to cause 65–75% of essential hypertension cases31. Obesity-related hypertension may be the result of one or more causes. It may lead to increased BP levels through activation of the renin–angiotensin–aldosterone system (RAAS), overactivation of the sympathetic nervous system, renal sodium reabsorption, and endothelial dysfunction, and increased insulin resistance32. These obesity indicators may be associated with visceral obesity. Previous studies in Asia found that WC and WHR were the most useful indicators for identifying adult diabetes and hypertension in South Asia33,34.

For WC, the rising trend of global obesity is significant. From 2013 to 2018, the average WC of men increased from 82 to 86.3 cm, and that of elderly women increased from 79.1 to 83.4 cm35. Most studies36 found a positive correlation between WC and BP, but not in a prospective study of a European population29. The differences in results may be due to ethnic differences, low participation rates, and relatively high withdrawal rates in the study. Previous studies in Iran found WC had a higher predictive value for hypertension than BMI in women aged 40–60, which was similar to our conclusion37. WHR and WC do not consider height. However, they can better reflect increases in visceral fat. Therefore, they may be preferred obesity indices of predicting hypertension in rural China. For BMI, given the significant differences in different regions, races, and body shapes, the use of a unified range may underestimate or overestimate the harm of the disease. Chinese people are relatively thin, and a previous study found that a BMI > 25 kg/m2 was associated with elevated BP10. Therefore, this study’s BMI standard (28 kg/m2) was relatively lower than European and American standards to avoid underestimation of obesity. Nevertheless, the predictive value of BMI for new-onset hypertension is relatively limited. This may be attributed to its inability to distinguish between muscle and fat; thus, it cannot reflect the actual degree of obesity, which is consistent with previous studies12.

In terms of subtle sex differences, menopause is a well-known risk factor for hypertension. The sudden decline in the circulating estrogen level38 may independently lead to high BP through some unknown mechanisms, such as the direct effect on the arterial wall, RAAS activation, and the sympathetic nervous system39. Considering that the baseline average age of women in the analysis cohort was 49.55 years, most of them had reached a menopausal state. The menopausal state was added as an adjustment factor in the female population. However, menopausal status and premature menopause (menopause < 45 years old) were not found to increase the risk of hypertension in this study cohort, which may be due to the relatively short follow-up time or relatively young age.

Our study has the following advantages. First, a prospective cohort study can fully demonstrate causality. Second, to the best of our knowledge, this is the first time a study has compared the predictive value of conventional and unconventional new obesity indicators for new-onset hypertension and explored the best predictive indicators in different sex and age groups; this will be conducive to the primary prevention of hypertension in primary medical and health institutions. Third, this study included a relatively comprehensive health screening program based on the general rural population, including routine blood biochemistry, physical measurement, diseases and demographic information, physical activities, and daily living habits. However, this study also has some limitations. First, the follow-up time was short. Second, our conclusions may only be applicable to the economically underdeveloped northeast rural areas in Asia. Due to the relatively thin body shape characteristics of Asians, the selection of predictive indicators may not be applicable to other races, and the extrapolation is limited. Third, given the low incidence of secondary hypertension in the general population, although the type of new-onset hypertension was further determined and verified in the follow-up (no suspicious secondary hypertension events were identified through the available information, such as renal function ion and symptom description), we did not use a gold standard, such as renal ultrasound, angiography, and respiratory and sleep detection.

Conclusions

In middle-aged and elderly populations, ABSI and BRI are independently and positively correlated with the risk of new-onset hypertension. Notably, different best predictors of hypertension exist in different age and sex subgroups. Women and men over the age of 60 should pay attention to WHR, and men under the age of 60 should pay attention to WC. Therefore, an in-depth subgroup analysis will help to design effective primary prevention strategies for specific middle-aged and elderly groups.